Explaining ethnic disparities in bachelor’s degree participation: evidence from NZ

ABSTRACT

There are substantial ethnic gaps in higher education in NZ, despite considerable policy effort aimed at this concern. This study uses newly linked administrative data to examine the underachievement of Māori and Pasifika relative to Europeans. We follow a population cohort born between 1990 and 1994 from school through to young adulthood to assess the relative contributions of prior academic performance, socioeconomic status, and parental education to these gaps. Controlling for the relevant covariates narrows the Māori-European gap, and eliminates the Pasifika-European gap in bachelor’s degree participation rates. Utilising Fairlie decompositions, we find that school performance is by far the largest contributor to the ethnic gaps. Socioeconomic status and parental education are also pertinent, but less important. Our results suggest that ethnic-based policies aimed at encouraging participation are likely to have a limited effect if used in isolation, and signal the need for policy interventions earlier in the education system.

KEYWORDS:

• Higher education
• ethnicity
• bachelor’s degree
• decomposition

JEL CLASSIFICATION:

• I21
• I24

Acknowledgements

We are grateful to several organisations for providing us with helpful comments and assistance with the data. This includes the Ministry of Education (Ralf Engler and David Earle), the Tertiary Education Commission (Anne Broadbent and Sergei Koudrin), the Treasury (Sarah Tumen), Motu (Isabelle Sin) and Statistics NZ’s microdata team. Thank you to Tim Maloney and Isabelle Sin for providing valuable review comments. We also thank the Productivity Commission’s ‘New models of tertiary education’ inquiry team for sponsoring this research and providing useful feedback. We also received helpful feedback from the participants of several workshops. Any errors or omissions remain the responsibility of the authors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 This under-representation is part of a wider trend in NZ of generally worse outcomes in a broad range of areas (such as income, health and housing) for Māori and Pasifika in comparison with other ethnic groups (particularly Europeans).

2 For instance, work by Broecke and Nicholls (2008) highlighted parental education as a key factor they would have liked to have included in their analysis of higher education outcomes in the UK, but was unfortunately not available in available data sources.

3 A meshblock is the smallest geographic unit used by Statistics NZ. See Section 4 for more details.

4 An exception to this is recent work by Ciao and Maloney (2017), which utilises Fairlie decompositions to examine ethnic differences in university grade outcomes and course completion rates.

5 For more information on the NCEA qualification see http://www.nzqa.govt.nz/qualifications-standards/qualifications/ncea/.

6 NZ has 8, predominantly government-funded universities, 18 polytechnics, and 3 wānanga. A wānanga is a tertiary education institution that provides programmes in a Māori cultural context.

7 Throughout this paper, observation counts are randomly rounded to base 3 in accordance with Statistics NZ confidentiality requirements.

8 Participation equates to being enrolled in a programme greater than 0.03 EFTS, which is a week of study.

9 Students entering a bachelor’s qualification at an NZ university before the age of 20 are generally required to have University Entrance or an equivalent qualification. However, universities offer special admission to those aged 20 or over who do not meet this criterion. See http://www.nzqa.govt.nz/qualifications-standards/awards/university-entrance/ and http://www.universitiesnz.ac.nzstudying-in-nz/domestic#Discretionary%20Entrance.

10 The July–June period roughly matches a school-intake cohort.

11 School is compulsory in NZ until the age of 16. It is also important to note that some NZ students study towards international qualifications (such as International Baccalaureate or Cambridge exams) instead of undertaking NCEA qualifications. These individuals will appear in the data on school enrolments, but the data do not allow us to identify students who undertake international qualifications. We, therefore, exclude students who attended schools that primarily offer international qualifications.

12 Due to the small nature of the ‘Other’ ethnic sub-group, we do not include descriptive statistics and results for this category.

13 The implication of using prioritised ethnicity is that multiple ethnic responses are not accounted for, thereby concealing diversity within, and overlap between, ethnic groups.

14 The deprivation index is constructed for each meshblock in NZ. A meshblock is the smallest geographic unit used by Statistics NZ. The median size of a meshblock in 2006 was approximately 87 people and around 35 households. In 2014 the median size of a high school catchment was 402 households.

15 Studies often use childhood cognitive test results as these are considered to be less influenced by environmental factors and, therefore, a better reflection of innate ability. For example, Maani (2006) uses IQ scores at age 8 and Chowdry et al. (2013) uses national assessed standardised achievement test scores at age 11. We do not have measures of childhood ability in our data, but use the earliest available measure (NCEA Level 1).

16 As the indigenous population, the percentage of Māori migrants may seem high. However, this may be because the NZ population is quite internationally mobile, with one of the largest diaspora in the world (in per capita terms) (Conway 2016). In particular, NZ and Australia have an open labour market and a substantial number of Māori live in Australia.

17 Indeed, state and state-integrated schools must offer NCEA, but can choose to offer other qualifications in addition to NCEA. Private schools are not obliged to offer NCEA.

18 Statistics NZ have collated information from several sources to identify parent–child links in the IDI, including birth records, Ministry of Social Development data relating to social welfare payments and family tax credits, Census 2013, and several household surveys.

19 We can obtain information on parental qualifications from the Census 2013 for 72% of our population cohort. While this sub-population is not random, it is broadly similar to our total population. The main difference is that this sub-population is less likely to include migrants.

20 We also test our results against those using estimated coefficients from a pooled regression that includes a dummy variable for the ethnic minority group (to account for the issue of the inappropriate transfer of some of the unexplained part of the gap into the explained component that arises with the use of the pooled method (see Jann 2008; Fortin 2006; and Fairlie 2005).

21 Usually, students attending state schools in NZ attend their local school. However, there is overlap in some school catchments, so students may live within more than one school catchment. In addition, if a school has extra places after all local students have enrolled, it can offer these places to students outside the catchment, with placements decided via a ballot.

22 For the participation decompositions, as a sensitivity test, we also used estimated coefficients from a pooled regression that included a dummy variable for the minority ethnic group. This made little difference to the results.

23 Note that the raw participation rates reported in Table 4 are slightly different to the descriptive statistics given in Table 2. This is because Table 2 descriptive statistics are based on the whole population, while Table 4 uses the sub-sample for which parents’ education was available (see Section 4).

24 Note that future research could delve further into the underlying causes of large disparities in school performance by NCEA Level 1 – there are likely a number of issues at play, such as intergenerational factors; attitudinal differences across educators; and potential racism in the education sector (see Gillborn 2010).

25 A few schools with very small numbers of students with no variation in outcomes were excluded from this analysis.